基于稀疏和低秩先验分离的快速动态MRI重建
陈思吉;杨晓梅;吕雪霜
【期刊名称】《计算机应用研究》 【年(卷),期】2016(033)010
【摘要】In order to accelerate dynamic MRI reconstruction,and pick out the dynamic organizations,this paper proposed a novel method based on separation via sparse plus low-rank prior.It separated the dynamic MRI into static backgrounds and dy-namic organizations by robust principal component analysis,built the corresponding low-rank matric and sparsity in x-f domain model,and then solved the optimization problem by alternating direction method of multipliers.Comparing with k-t FOCUSS algorithm and k-t SLR algorithm,this method can ensure the image quality on PSNR,and structural similarity(SSIM).The experimental results show that this method significantly improves imaging speed,reduces the motion artifacts,and stands out the dynamic information.%为了加速动态核磁共振成像的重建，并提取动态组织部分，提出一种基于将稀疏和低秩先验分离的重建方法。算法利用鲁棒主成分分析法，将动态MRI看做静态背景和动态组织的合成，建立相应的低秩矩阵和x-f域稀疏模型，再通过交替方向拉格朗日乘子法求解优化问题。与经典的k-t FOCUSS算法和k-t SLR算法进行对比，提出的算法能保证重建质量，即峰值信噪比（PSNR）、结构相似性（SSIM）等评价指标。实验结果表明，该算法能实现快速动态MRI的成像，减少运动伪影，同时更利于提取动态信息。
基于稀疏和低秩先验分离的快速动态MRI重建
